A framework for end-to-end RSVP context transfer in Mobile IPv6(MIPv6) based on the architecture of F-HMIPv6 is proposed in this paper. The scheme provides an end-to-end RSVP context transfer for real-time application...A framework for end-to-end RSVP context transfer in Mobile IPv6(MIPv6) based on the architecture of F-HMIPv6 is proposed in this paper. The scheme provides an end-to-end RSVP context transfer for real-time applications to overcome the weakness of Context Transfer Protocol which can not meet the need of end-to-end QoS mechanisms because contexts are only transferred between Access Routers(ARs), therefore they can promptly get the same forwarding process, minimize the handover service disruption, and avoid initiating the end-to-end RSVP signaling from scratch after an MN performs handovers. It may also reduce the signaling overhead and handover latencies by adopting the F-HMIPv6architecture. The performance of the approach is compared with the re-initiating RSVP signaling to re-establish QoS states using network simulator, and the numerical results show that the scheme has the less latency and packet loss than that of the re-initiating approach.展开更多
Mobility support for the next generation IPv6 networks has been one of the recent research issues due to the growing demand for wireless services over internet.In the other hand,3GPP has introduced IP Multimedia Subsy...Mobility support for the next generation IPv6 networks has been one of the recent research issues due to the growing demand for wireless services over internet.In the other hand,3GPP has introduced IP Multimedia Subsystem as the next generation IP based infrastructure for wireless and wired multimedia services.In this paper we present two context transfer mechanisms based on predictive and reactive schemes,to support seamless handover in IMS over Mobile IPv6.Those schemes reduce handover latency by transferring appropriate session information between the old and the new access networks.Moreover,we present two methods for QoS parameters negotiations to preserve service quality along the mobile user movement path.The performances of the proposed mechanisms are evaluated by simulations.展开更多
According to Edward Hall,cultures can be categorized as being either high or low context.Chinese culture is typically high-context culture.The paper tries to explore the mode of information transfer in high-context cu...According to Edward Hall,cultures can be categorized as being either high or low context.Chinese culture is typically high-context culture.The paper tries to explore the mode of information transfer in high-context cultures in terms of Red Cliff.The discussion will be focused on dialogues and the costume.This will help understand the characteristics of high-context culture and make the communication between people from cultures of different contexts easier.展开更多
大语言模型(large language models,LLMs)在知识存储方面不断增强的能力展示了其作为知识库的潜在效用.然而,任何给定的提示只能提供大语言模型所涵盖知识的下限估计.在语言模型即知识库(language models as knowledge bases,LMs-as-KBs...大语言模型(large language models,LLMs)在知识存储方面不断增强的能力展示了其作为知识库的潜在效用.然而,任何给定的提示只能提供大语言模型所涵盖知识的下限估计.在语言模型即知识库(language models as knowledge bases,LMs-as-KBs)的场景中,先前的提示学习方法忽略了查询风格对模型表现的影响.揭示了大语言模型确实具有与查询风格相关的可学习偏好,并且利用大语言模型的这种特性引入了查询风格自适应转换(adaptive query style transfer,ARES)方法,通过适应大语言模型的偏好来增强其知识查询的表现. ARES方法从构造查询候选集开始,通过改写实现多种表达风格的纳入.随后,ARES训练一个评估器来学习并识别大语言模型对查询风格的偏好,评估查询候选集并选择潜在最优查询.在多个数据集上进行的实验表明了该方法在提高大语言模型即知识库服务上查询准确率的有效性,增量对比原始模型与3个基线方法分别实现了平均2.26个百分点,1.68个百分点,1.19个百分点,1.17个百分点的提升,这表明ARES可以与其他方法有效地结合使用,从而实现多角度的查询表现增强.展开更多
针对野生动物数据集样本量小、目标尺度多变所导致的野生动物检测困难以及检测精度低等问题,提出一种基于多尺度上下文提取的小样本野生动物检测(MS-FSWD)算法。首先,通过多尺度上下文提取模块增强模型对不同尺度的野生动物的感知能力,...针对野生动物数据集样本量小、目标尺度多变所导致的野生动物检测困难以及检测精度低等问题,提出一种基于多尺度上下文提取的小样本野生动物检测(MS-FSWD)算法。首先,通过多尺度上下文提取模块增强模型对不同尺度的野生动物的感知能力,提高检测性能;其次,引入Res2Net作为原型校准模块的强分类网络对分类器输出的分类分数进行校正;然后,在RPN中加入置换注意力机制,增强目标区域的特征图,弱化背景信息;最后,将平衡L1损失作为定位损失函数,提升目标定位性能。实验结果表明,相比DeFRCN算法,MS-FSWD在小样本野生动物数据集FSWA上,1-shot和3-shot检测任务中新类AP50分别提升了9.9%和6.6%;在公共数据集PASCAL VOC上,MS-FSWD最高提升了12.6%。与VFA算法相比,在PASCAL VOC数据集Novel Set 3的10-shot任务中,新类AP50提升了3.3%。展开更多
随着基于位置的社交网络的快速发展,下一个PoI(point of interest)推荐已成为推荐领域的研究热点。然而现有研究模型忽略了PoI的时空特征以及上下文信息对下一个PoI推荐的效果。针对该问题,提出一种时空上下文感知的下一个PoI推荐方法...随着基于位置的社交网络的快速发展,下一个PoI(point of interest)推荐已成为推荐领域的研究热点。然而现有研究模型忽略了PoI的时空特征以及上下文信息对下一个PoI推荐的效果。针对该问题,提出一种时空上下文感知的下一个PoI推荐方法。首先,利用图注意力网络(GAT)学习包含社交关系的用户表征;并且通过流行度增强二部图神经网络(PEBGNN)学习含有PoI交互偏好的用户表征和PoI表征;同时,利用时空图卷积网络(ST-GCN)学习PoI时空转移偏好的PoI表征;最后,通过融合所学到的用户表征和PoI表征,计算出用户对于各个PoI的预测评分,以此为基础为用户推荐下一个PoI。为了验证该方法的有效性,在Gowalla、Foursquare以及Yelp这三个公开的数据集上进行了测试。实验结果显示,相比于多个基准模型,所提方法在准确率和召回率方面均展现出了显著的优势,分别平均提升28.53%和7.65%。展开更多
文摘A framework for end-to-end RSVP context transfer in Mobile IPv6(MIPv6) based on the architecture of F-HMIPv6 is proposed in this paper. The scheme provides an end-to-end RSVP context transfer for real-time applications to overcome the weakness of Context Transfer Protocol which can not meet the need of end-to-end QoS mechanisms because contexts are only transferred between Access Routers(ARs), therefore they can promptly get the same forwarding process, minimize the handover service disruption, and avoid initiating the end-to-end RSVP signaling from scratch after an MN performs handovers. It may also reduce the signaling overhead and handover latencies by adopting the F-HMIPv6architecture. The performance of the approach is compared with the re-initiating RSVP signaling to re-establish QoS states using network simulator, and the numerical results show that the scheme has the less latency and packet loss than that of the re-initiating approach.
文摘Mobility support for the next generation IPv6 networks has been one of the recent research issues due to the growing demand for wireless services over internet.In the other hand,3GPP has introduced IP Multimedia Subsystem as the next generation IP based infrastructure for wireless and wired multimedia services.In this paper we present two context transfer mechanisms based on predictive and reactive schemes,to support seamless handover in IMS over Mobile IPv6.Those schemes reduce handover latency by transferring appropriate session information between the old and the new access networks.Moreover,we present two methods for QoS parameters negotiations to preserve service quality along the mobile user movement path.The performances of the proposed mechanisms are evaluated by simulations.
文摘According to Edward Hall,cultures can be categorized as being either high or low context.Chinese culture is typically high-context culture.The paper tries to explore the mode of information transfer in high-context cultures in terms of Red Cliff.The discussion will be focused on dialogues and the costume.This will help understand the characteristics of high-context culture and make the communication between people from cultures of different contexts easier.
文摘针对野生动物数据集样本量小、目标尺度多变所导致的野生动物检测困难以及检测精度低等问题,提出一种基于多尺度上下文提取的小样本野生动物检测(MS-FSWD)算法。首先,通过多尺度上下文提取模块增强模型对不同尺度的野生动物的感知能力,提高检测性能;其次,引入Res2Net作为原型校准模块的强分类网络对分类器输出的分类分数进行校正;然后,在RPN中加入置换注意力机制,增强目标区域的特征图,弱化背景信息;最后,将平衡L1损失作为定位损失函数,提升目标定位性能。实验结果表明,相比DeFRCN算法,MS-FSWD在小样本野生动物数据集FSWA上,1-shot和3-shot检测任务中新类AP50分别提升了9.9%和6.6%;在公共数据集PASCAL VOC上,MS-FSWD最高提升了12.6%。与VFA算法相比,在PASCAL VOC数据集Novel Set 3的10-shot任务中,新类AP50提升了3.3%。
文摘随着基于位置的社交网络的快速发展,下一个PoI(point of interest)推荐已成为推荐领域的研究热点。然而现有研究模型忽略了PoI的时空特征以及上下文信息对下一个PoI推荐的效果。针对该问题,提出一种时空上下文感知的下一个PoI推荐方法。首先,利用图注意力网络(GAT)学习包含社交关系的用户表征;并且通过流行度增强二部图神经网络(PEBGNN)学习含有PoI交互偏好的用户表征和PoI表征;同时,利用时空图卷积网络(ST-GCN)学习PoI时空转移偏好的PoI表征;最后,通过融合所学到的用户表征和PoI表征,计算出用户对于各个PoI的预测评分,以此为基础为用户推荐下一个PoI。为了验证该方法的有效性,在Gowalla、Foursquare以及Yelp这三个公开的数据集上进行了测试。实验结果显示,相比于多个基准模型,所提方法在准确率和召回率方面均展现出了显著的优势,分别平均提升28.53%和7.65%。
